Abstract
Vascular endothelial growth factor (VEGF) is a major inducer of angiogenesis. We generated a transgenic reporter mouse, VEGF-GL, in which an enhanced green fluorescent protein-luciferase fusion protein is expressed under the control of a human VEGF-A promoter. The VEGF-GL mouse exhibited intense bioluminescence throughout the body at 1 week of age. The signals rapidly declined to a relatively low level as the mice grew. The adult VEGF-GL mouse showed restricted bioluminescence to the areas undergoing wound healing. In contrast, the VEGF-GL mice, which were crossed with mouse mammary tumor virus-polyoma virus middle T antigen transgenic mammary tumor mice, exhibited prominent bioluminescence in the tumors, correlating with VEGF transcription. Tumor bioluminescence was observed in the bigenic mice as early as 8 weeks, before tumors were palpable, and the signals increased with tumor growth. In conclusion, the VEGF-GL mouse permits longitudinal and quantitative assessment of VEGF promoter activity in vivo. The model should facilitate understanding of the molecular controls and pathways that regulate VEGF transcription in vivo.
VEGF expression is transcriptionally regulated by various stimuli. Hypoxia is the most potent inducer of VEGF transcription. 5 Hypoxia owing to rapid tissue growth or ischemia upregulates hypoxia-inducible factor 1α (HIF-1α), a key mediator of the hypoxic cellular responses, and the binding of HIF-1α to the hypoxia-responsive element (HRE) in VEGF promoter potently induces VEGF transcription.6,7 VEGF transcription is also upregulated by growth factors, inflammatory cytokines, and hormones, including epidermal growth factor, heregulin, platelet-derived growth factor, transforming growth factor β1, tumor necrosis factor α, interleukin-1β, and estrogen. 5 This suggests that paracrine and autocrine release of such factors cooperates with local hypoxia to upregulate VEGF transcription in the tissue microenvironment. Indeed, recent studies have demonstrated microenvironment-dependent or tumor cell type-specific regulation of VEGF transcription.8–10 An animal model that permits spatiotemporal monitoring of VEGF transcription in the natural microenvironment undoubtedly improves understanding of VEGF gene activation in vivo.
Among newly developed in vivo analysis systems, bioluminescence imaging (BLI) provides a unique opportunity to assess molecular and cellular events in living animals. 11 The system uses the chemiluminescent oxidative reaction between a firefly luciferase reporter enzyme and its substrate luciferin in the presence of oxygen, adenosine triphosphate, and Mg2+, producing broadband (500-700 nm) light with a peak emission at 563 nm. The properties of luciferase-based light emission allows noninvasive, sensitive, and real-time molecular imaging in living organisms owing to the fact that the wavelengths above 600 nm are only weakly absorbed by endogenous hemoglobin. This feature has enabled quantitative imaging of internal organs, a major limitation of fluorescence imaging. BLI provides an ideal system for longitudinal monitoring of a disease course in the same animal.
In the present study, we generated a transgenic reporter mouse, VEGF-GL, in which VEGF promoter activity can be assessed by BLI in vivo. By crossing VEGF-GL mice with mouse mammary tumor virus-polyoma virus middle T antigen (MMTV-PyVmT) mammary tumor mice, we assessed VEGF promoter activity in murine mammary tumorigenesis. Our results demonstrate that VEGF induction during tumor growth can be monitored by the VEGF-GL model. The model should facilitate understanding of the molecular pathways that regulate VEGF transcription in vivo.
Materials and Methods
Transgene Construct and Generation of Transgenic Mouse
The VEGF-GL (VEGF-enhanced green fluorescent protein [E
In Vivo BLI
Images were taken using a nitrogen-cooled charge-coupled device (CCD) camera (EB-1300, Roper Scientific, Tucson, AZ) coupled to a light-tight imaging chamber. Mice were anesthetized with a mixture of ketamine and xylazine (50 mg/kg, 10 mg/kg, intraperitoneally), and an aqueous solution of
Mammary Tumor Study
A transgenic mammary tumor was induced in the VEGF-GL mice by crossing the mice with FVB MMTV-PyVmT mice (Jackson Laboratory, Bar Harbor, ME), expressing PyVmT under the MMTV long terminal repeat (LTR). 14 The bigenic female mice were imaged weekly for VEGF bioluminescence from the age of 7 weeks. Tumor size was measured externally using calipers, and tumor volume was calculated as previously described. 15
Immunoprecipitation and Western Blotting
Reporter protein expression was examined by immunoprecipitation and Western blot analysis. Mammary tumor tissues were removed from 13-week-old bigenic (VEGF-GL/MMTV-PyVmT) mice. A 10% tissue lysate (wt/vol) was prepared by homogenization of the tissues in lysis buffer (50 mM Tris/pH. 7.5, 150 mM NaCl, 1% Triton X-100, 0.5% deoxycholate, 1 mM ethylenediaminetetraacetic acid [EDTA]) containing a protease inhibitor cocktail (Complete Mini, Roche Applied Science, Indianapolis, IN) and two subsequent centrifugations (15,000g for 30 minutes). Twenty milligrams of tissue lysate was subjected to immunoprecipitation with anti-GFP monoclonal antibody (Roche Applied Science). Bound proteins were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis on 7.5% gels under a reducing condition and subjected to blotting with anti-luciferase antibody (Cortex Biochem, San Leandro, CA).
In Situ Hybridization
Spatial colocalization of VEGF and luciferase transcription was determined by in situ hybridization. Mammary tumor tissues were sampled from 13-week-old bigenic mice, and in situ hybridization was performed essentially as described previously. 16 Template complementary deoxyribonucleic acids (cDNAs) for riboprobes (mouse VEGF, 250 bp; firefly luciferase, 400 bp) were generated by PCR and subcloned into a pCR4-TOPO vector (Invitrogen, Carlsbad, CA). Background hybridization was assessed using the sense probe.
Mammary Tumor Cell Isolation and Hypoxic Cell Culture Experiment
Hypoxia-induced reporter gene induction in isolated bigenic mammary tumor cells was examined. Tumors from 14-week-old bigenic females were minced in 10% fetal bovine serum (FBS)-Dulbecco's Modified Eagle's Medium (DMEM), washed with phosphate buffered saline (PBS), and digested at 37°C for 2 hours in 0.25% trypsin and 2 mM EDTA in PBS. After washing with PBS, the cell suspension was plated on plastic dishes with 10% FBS-DMEM, and the grown cells were subcloned and characterized by immunostaining and Western blot using antibodies to E-cadherin (BD Transduction Laboratories, San Jose, CA), cytokeratins (pancytokeratin mix, Sigma, St. Louis, MO), and vimentin (Sigma). Mammary epithelial tumor cells expressing both E-cadherin and cytokeratin were used for the study. For a hypoxic cell culture study, 5 × 105 tumor cells were plated in six-well plates (Falcon) and starved in 1% FBS medium overnight. Cells were placed in 0.2% oxygen and harvested at 0, 6, 12, and 24 hours for reverse transcription polymerase chain reaction (RT-PCR) analysis and luciferase activity assay. Luciferase activity assay of the cell lysates was performed according to the manufacturer's instructions (Promega, Madison, WI).
Reverse Transcription Polymerase Chain Reaction
VEGF and luciferase transcription in isolated tumor cells was evaluated by RT-PCR. Ribonucleic acid (RNA) was isolated using RNA STAT-60 (TEL-TEST, Inc., Friendwood, TX), and mRNA was purified from the RNA using the Oligotex mRNA mini kit (Qiagen, Valencia, CA). Single-stranded cDNA was synthesized using the cDNA cycle kit (Invitrogen). PCRs were performed with the specific primers for firefly luciferase (5′-CTCTGGCACCTAATCCAG-3′, 5′-GCTGATGTA GTCTCAGTG-3′), mouse VEGF (5′-TGAACTTTCTGCTCTCTTGG-3′, 5′-CTCCAGGGCTTCATCGTT-3′, 5′-AGTGATCAAGTTCATGGA-3′), and mouse β-actin (5′-GTCGTACCACAGGCATTGTGATGG-3′, 5′-GCAATGCCTGGGTACATGGTGG-3′).
PCR amplification products were separated by 1% agarose gel, visualized with ethidium bromide, photographed, and semiquantified using AlphaImager software (AlphaInnotech Corp., San Leandro, CA).
Results
Generation of the VEGF-GL Transgenic Mouse
To visualize VEGF promoter activity by fluorescence microscopy and BLI, an EGFP-luciferase fusion gene 17 was inserted downstream of a human VEGF-A promoter 12 (Figure 1A). Of 38 F0 mice, two transgenic lines, VEGF-GL #31 and VEGF-GL #33, were identified by PCR. BLI of these two lines showed different levels of luciferase expression. As shown in Figure 1B, the VEGF-GL #31 line displayed significant luciferase signals only in sites of wounding, such as the ear and tail clip, and physiologic angiogenesis, such as female reproductive cycle and incisor teeth growth (data not shown), when the sensitivity is increased. On the other hand, the VEGF-GL #33 line exhibited bright bioluminescence throughout the body (data not shown). The copy number of transgenes was 2 in the VEGF-GL #31 line and 10 in the VEGF-GL #33 line (data not shown). Given that low background signal is desirable to assess VEGF gene activation, the VEGF-GL #31 mouse was further characterized. As shown in Figure 1C, the VEGF-GL #31 mice revealed intense bioluminescence throughout their bodies at the age of 1 week, presumably owing to growth-related tissue hypoxia. RT-PCR analysis detected the EGFP-luciferase transcripts in various organs of the 1-week-old VEGF-GL mice, including the brain, lung, liver, spleen, intestines, and skeletal muscle (data not shown). As anticipated, by 4 weeks of age, the mice reached near-adult size and the basal luciferase signal was rapidly reduced to a relatively low level (see Figure 1C). Although VEGF is highly expressed in the kidney and heart, significant reporter expression was not detected in these organs either by RT-PCR or tissue luciferase activity assay (data not shown).
Generation of a vascular endothelial growth factor (VEGF)-GL transgenic mouse. A, Schematic representation of the VEGF-GL transgene construct. The 2,849 bp of the human VEGF-A promoter 12 was linked to a fusion gene of enhanced green fluorescent protein (EGFP) and luciferase. The number indicates the nucleotide position from the transcriptional start site as +1. The sets of the polymerase chain reaction primers (P1, P2) used for genotyping are also shown. AP-1, AP-2, and SP-1 = the binding sites of activator proteins 1 and 2 (AP-1, AP-2) and simian virus 40 promoter factor 1 (SP-1); E = EcoR I; HRE = hypoxia-responsive element; 3’ HS1 = human globin 3’ DNaseI hypersensitive site; IgG1 = immunoglobulin G1; N = Not I; S = Spe I. B, Bioluminescence imaging (BLI) of an adult VEGF-GL #31 mouse and a wild-type littermate. The color map represents the photon counts per second. Significant bioluminescence is observed in the ear punch and tail clip sites (arrows) of the VEGF-GL mouse. C, BLI of postnatal VEGF-GL mice. VEGF-GL mice at postnatal days 7 (P7), 10 (P10), and 14 (P14) were imaged, and whole-body luciferase signals (closed circles) were quantified. Data are presented as mean ± standard error of measurement. The photon counts of nontransgenic littermates (open circles; n = 3) are also plotted.
VEGF-GL Expression in MMTV-PyVmT Transgenic Mammary Tumors
Experimental and clinical data have shown that breast cancer is an angiogenesis-dependent disease associated with VEGF expression.18,19 To further characterize the VEGF-GL #31 (hereafter referred to as VEGF-GL) mouse, we induced transgenic mammary tumors in the VEGF-GL mice by crossing the mice with MMTV-PyVmT mice. As shown in Figure 2A, aggressive mammary tumors were developed in the VEGF-GL mice carrying the MMTV-PyVmT transgene, and the bigenic mice displayed prominent tumor bioluminescence. Consistent with the results of BLI, immunoprecipitation and Western blot analysis using anti-GFP and anti-luciferase antibodies detected a predicted size (86 kDa) of EGFP-luciferase protein in the bigenic tumor tissues (see Figure 2A, right panel), confirming the presence of the transgene in the tumor tissue.
To assess spatial correlation of VEGF-GL expression and endogenous VEGF transcription, in situ hybridization was performed on the MMTV-PyVmT/VEGF-GL and MMTV-PyVmT tumor tissue sections. As shown in Figure 2B, luciferase mRNA was detected in the tumor nodules of MMTV-PyVmT/VEGF-GL mice, correlating with VEGF mRNA expression, wherease no luciferase transcript was detected in MMTV-PyVmT tumors. We also assessed colocalization of the reporter protein and endogenous VEGF by fluorescence microscopy and immunohistochemistry; however, low-level expression of the reporter made it difficult to evaluate their colocalization at the cellular level.
Further, we evaluated activation of the VEGF-GL transgene to hypoxia exposure, a well-known inducer of VEGF transcription, in the mammary tumor cells (Figure 2C). Mammary epithelial tumor cells were isolated from MMTV-PyVmT/VEGF-GL and MMTV-PyVmT mice. The cells were exposed to 0.2% oxygen for up to 24 hours. VEGF and luciferase mRNA levels were examined by RT-PCR and semiquantified. As shown in Figure 2C, luciferase mRNA increased fourfold in the first 12 hours, with an overall fivefold increase after 24 hours in MMTV-PyVmT/VEGF-GL tumor cells, and this correlated with endogenous VEGF mRNA induction. The luciferase activity was increased in a manner similar to luciferase transcription (see Figure 2C). Luciferase transcription and activity were undetectable in MMTV-PyVmT tumor cells lacking the VEGF-GL transgene (data not shown). Low pH (pH 6.6) and hydrogen peroxide (1 mM) exposure also increased luciferase activity in the bigenic tumor cells 3- and 4.5-fold (at 8 hours), as reported previously.20,21
Vascular endothelial growth factor (VEGF)-GL expression in a mouse mammary tumor virus-polyoma virus middle T antigen (MMTV-PyVmT) transgenic mammary tumor. A, Bioluminescence imaging (BLI) of 13-week-old female mice from VEGF-GL X MMTV-PyVmT crosses (left). Prominent tumor bioluminescence is noted in the bigenic mouse. Enhanced green fluorescent protein (EGFP)-luciferase reporter protein was immunoprecipitated from mammary tumor extracts using anti-GFP antibody and immunoblotted with anti-luciferase (anti-Luc) antibody. An arrow indicates an 86 kDa EGFP-luciferase protein. B, VEGF and luciferase transcription in mammary tumor tissues. In situ hybridization was performed on serial sections of VEGF-GL/MMTV-PyVmT and MMTV-PyVmT tumors using VEGF and luciferase probes. The signals were indicated by red color. Spatially colocalized expression of VEGF and luciferase in a VEGF-GL/MMTV tumor nodule (asterisk) (hematoxylin counterstain; X200 original magnification). C, VEGF and luciferase induction in bigenic mammary tumor cells under hypoxic exposure. Mammary tumor cells were isolated from VEGF-GL/MMTV-PyVmT mice and exposed to 0.2% oxygen. Luciferase and endogenous VEGF (mouse VEGF) induction was determined by reverse transcription (RT)-polymerase chain reaction (upper panel). Luciferase induction was also assessed by luminometer assay (lower panel). The values indicate mean ± standard error of measurement of three independent experiments.
Longitudinal Monitoring of VEGF Promoter Activity in Mammary Tumor Growth
The bigenic (MMTV-PyVmT/VEGF-GL) mice were imaged weekly for VEGF bioluminescence, and the tumor size (volume) was calculated externally using caliper measurements. Palpable mammary tumors were developed in the bigenic mice between 9 and 12 weeks of age. As shown in Figure 3A, significant bioluminescence was detected in the mammary regions of the bigenic mice as early as 8 weeks of age, when no palpable tumors were present. BLI and photon count analysis showed increasing and extending luciferase signals on tumor growth (see Figure 3, A and B). Regional photon count measurements revealed a different time course expression of luciferase in individual mammary glands (see Figure 3B, right panel). A significant correlation was observed between tumor photon counts and the tumor volume measured externally in a single animal over time or on excised tumors (data not shown). Although intense luciferase signals were observed in the bigenic mammary tumors, no GFP signal was detected in the tumor tissues or cells by confocal microscopy and flow cytometry (data not shown).
Longitudinal assessment of vascular endothelial growth factor (VEGF) promoter activity in murine mammary tumorigenesis. A, Repetitive bioluminescence imaging of VEGF-GL/mouse mammary tumor virus-polyoma virus middle T antigen (MMTV-PyVmT) mice. The bigenic females were imaged from 7 to 13 weeks of age. A representative mouse is shown. B, Photon counts of the whole body (left; n = 5) and individual mammary glands (right, a single animal) were quantified with VEGF-GL/MMTV-PyVmT females over a 7-week period. The values indicate mean ± standard error of measurement. The photon counts of nontransgenic females, lacking both VEGF-GL and MMTV-PyVmT transgenes, are also plotted (open circles; n = 3).
Discussion
We generated a VEGF-GL reporter mouse. The 1-week-old VEGF-GL mice displayed intense bioluminescence. The signals declined rapidly as the mice grew, consistent with a previous report showing remarkable downregulation of VEGF transcription during postnatal organ development. 22 The adult VEGF-GL mouse showed a very low level of bioluminescence. In contrast, the VEGF-GL/MMTV-PyVmT mice showed prominent bioluminescence in their mammary regions. Tumor bioluminescence was observed in the VEGF-GL/MMTV-PyVmT mice as early as 8 weeks, before tumors were palpable. It is known that tumor cells activate quiescent endothelial cells at the premalignant stage through a process termed “angiogenic switch,” which is associated with VEGF expression. 23 In addition, Lin and colleagues reported that premalignant lesions are observed in MMTV-PyVmT mice at 8 to 9 weeks prior to the angiogenic response at 9 to 12 weeks. 24 Taken together, our data demonstrate that BLI provides a highly sensitive in vivo assay that senses VEGF induction in early-stage mammary tumors.
VEGF bioluminescence in mammary regions or in a solid tumor mass increased during the entire tumor growth. The results demonstrate that VEGF promoter activity in tumor growth can be longitudinally assessed by BLI in vivo. Indeed, a recent study showed that exponential growth of luciferase-expressing tumor cells can be visualized by BLI over 28 days in vivo. 25 It is noteworthy that VEGF bioluminescence does not increase exponentially on tumor growth. This could be due to unequal VEGF mRNA expression in the tumors. In situ hybridization studies have shown that the VEGF mRNA levels are variable within a mammary tumor. VEGF mRNA expression is occasionally greater in the center in large islands of invasive breast tumor cells or in viable tumor cells adjacent to necrotic foci.26,27 Also, it is conceivable that VEGF promoter activity is downregulated when the solid tumor acquires the necessary vasculature to sustain itself.
In this study, we used a fusion protein of EGFP and firefly luciferase as a reporter. However, GFP fluorescence was undetectable in our model even in the tumor tissues. This is likely due to both the low level of transgene expression and lower sensitivity of GFP detection. During the present study, we generated Chinese hamster ovary (CHO) cell lines stably expressing GFP-luciferase fusion protein. In this experiment, we also observed prominent luciferase, but not GFP, signals in the clones expressing low levels of GFP-luciferase protein. In addition, recent studies using the tumor cells expressing both GFP and luciferase proteins have shown that BLI is a more sensitive method than fluorescence imaging because its background is minimal.15,28 We think it is unlikely that the GFP signal is reduced by its fusion to luciferase because fluorescence-activated cell-sorting analysis of the CHO cells transfected with GFP or GFP-luciferase plasmids exhibited comparable GFP fluorescence. Low levels of the reporter expression in the VEGF-GL mouse were also evidenced by undetectable transgene expression in Northern blot analysis using isolated tissue mRNA, Western blot analysis with crude tissue lysates, and anti-GFP or anti-luciferase immunohistochemistry (data not shown). Indeed, quantification of a reporter protein using recombinant GFP protein as standards revealed low levels of EGFP-luciferase expression (< 1.0 ng/mg lysate) in the bigenic mammary tumor tissues. This is a weakness of this model; however, it is also a merit of this model because VEGF is expressed in various tissues and a high level of reporter expression may cause high background signals. In fact, we could not assess VEGF gene activation in the transgenic line mouse (VEGF-GL #33) expressing a high level of the reporter. Fusion of luciferase with a more sensitive reporter enzyme, such as β-galactosidase, would be a future direction for dual characterization of VEGF promoter activity.
In the present study, we used a human VEGF promoter. As demonstrated by a recent study of the mice expressing a humanized form of VEGF-A, 29 it would be important to assess how the human VEGF promoter works in mouse tumor models in vivo; however, this would not be a big concern in this model because the human VEGF promoter is structurally very similar to the mouse VEGF promoter. 30 In addition, the human VEGF promoter region used in this study was shown to contain critical elements for VEGF gene induction, including HRE, AP-1, and SP-1/SP-3, and responds to various known VEGF inducers when it is introduced to cultured mammalian cells, as shown in Figure 2C.20,21 Although we could not observe the reporter expression in the heart and kidney in which VEGF is abundantly expressed, this could be due to an incomplete promoter sequence, lacking a tissue-specific regulatory element for these organs, because other transgenic experiments using the same or a slightly different promoter region also failed to generate a significant reporter signal in these tissue sites.12,31 Because heterozygous VEGF gene inactivation was shown to cause vascular defects,32,33 further characterization of the VEGF promoter region than the luciferase knock-in strategy would be required to resolve this problem.
The present study demonstrated that VEGF gene activation in a murine tumor model can be assessed in vivo in real time using BLI. This system offers a unique opportunity to obtain considerable temporal information from a single animal. Neoplastic disease is a complex, multistep process involving the loss of genetic and immune regulation and interactions between the transformed cells and surrounding tissues. Given that various factors have been shown to affect VEGF transcription, 5 it is conceivable that VEGF transcription is differentially regulated in the tumors, depending on stage, cell type, and tissue microenvironment. Indeed, recent studies have demonstrated microenvironment-dependent regulation of tumor VEGF expression.8,34 The VEGF-GL mouse, together with murine tumor models, provides an important in vivo assay to investigate VEGF gene regulation during tumorigenesis in the natural microenvironment. The model would be a powerful tool to elucidate the molecular pathway triggering VEGF transcription in mammary tumorigenesis in vivo when it is combined with specific inhibitor treatments or knockout mice. Further, it has been shown that tumor-associated stroma is an important source of VEGF production.12,35 The VEGF-GL reporter mouse could be an extremely useful tool to assess stromal VEGF transcription when used in conjunction with a tumor cell implantation assay. Finally, noninvasive in vivo imaging facilitates the use of fewer animals, rapid and quantitative assessment, and better control of the experiment, especially for those in which temporal changes are of great importance and interest. The VEGF-GL model could also be a powerful tool for evaluating anti-VEGF effects of pharmacologic agents or for testing new antiangiogenesis protocols. Thus, a VEGF-GL reporter mouse offers a unique opportunity to investigate VEGF gene activation in living animals. This model should facilitate understanding of molecular controls that regulate VEGF gene activation in vivo.
Located in the heart of the Texas Medical Center, M. D. Anderson's Houston campus covers 18 acres with 13 facilities, and has served patients from every state in the nation and many foreign countries. The M. D. Anderson family has more than 16,000 members, each playing a critical role in our mission to eliminate cancer - and each responsible for our worldwide reputation as a center of excellence, cutting-edge care and new therapies. We're proud of our team and invite you to join us in Houston and be a part of Making Cancer History TM .
Houston, America's fourth largest city, offers diverse and cultural activities including a downtown Theater District, a world-renowned Museum District, world-class shopping, and many recreational parks and professional sports facilities.
The Department of Experimental Diagnostic Imaging is currently accepting applications for an Associate Professor, Tenure Track appointment.
Doctoral degree from an accredited institution of higher education in Chemistry, Chemical Physics, or related discipline with published experience Magnetic Resonance Spectroscopy is required. Familiarity with investigation of cellular metabolism in normal and transformed models, as well as apoptotic death and response to novel targeted cancer therapies is desired. Preference will be given to individuals who also have experience with in vivo MRS in cancer. Evidence of success in obtaining funding for molecular imaging research is expected. Record of publications in quality peer reviewed journals.
Interested candidates are invited to contact:
M. D. Anderson Cancer Center is an EOE employer and does not discriminate on the basis of race, color, national origin, gender, sexual orientation, age, religion, disability or veteran status, except where such distinction is required by law. All positions at The University of Texas M. D. Anderson Cancer Center are security-sensitive and subject to Texas Education Code 51.215, which authorizes the employer to obtain criminal history record information. Smoke-free and drug-free environment.
Footnotes
Acknowledgments
We thank Dr. Brian Seed for providing the VEGF-GFP construct; Drs. Lynn Matrisian, Tom Daniel, Tsutomu Kume, and Scott Baldwin for supports; and Sabrina Hannah and Amy Hsu for technical assistance. We also thank the Vanderbilt-Ingram Cancer Center Transgenic/ES Cell Facility.